> Can you imagine every piece of software checking for every piece of esoteric hardware and having some special behavior?

This is what MS Office does. A Microsoft engineer told me once that because many printers lie to the OS about their capabilities, they built a giant lookup table into the Office, that basically consists of "if printer model X, vendor Y and firmware Z, then don't trust it about the fonts onboard, render the text before sending instead".

I remember back in the day doing DirectDraw/Direct3D programming, I had to have lots of "if video card type == x, then ignore these capabilities and do this other thing instead" hacks. A much bigger issue back then when we had more than three video card manufacturers. (it's still a minor issue today ... nVidia cards tend to be way more permissive about OpenGL calls than AMD cards. But I can afford two discrete cards and one onboard chipset.)

It's so completely impractical for a hobbyist project to own every possible piece of hardware, or at least a huge majority of it. Yet it's nothing for a Fortune 500 company to do the same.

Really slants the playing field away from independent software developers.

Or, you could crowdsource an open-source abstraction layer; this would level the playing field again.

Impractical and impossible are of course two different things. It's sort of funny to consider a scenario in which a manufacturer adds a thermistor to the speaker coil assembly to measure the temperature & lower the amplifier output via the BIOS as the temperature in the voice coil rises to a dangerous level. I suspect that might cost more than using good speakers in the first place but then again if there's limited room for speakers which would lead to heat buildup in the voice coil such a solution might be practical.